AUTOMATED NON-INVASIVE DETERMINATION OF THE SEX OF AN EMBRYO AND THE FERTILITY OF A BIRD EGG.

专利摘要:
a method of automated non-invasive sex determination of a bird egg embryo (14) is disclosed herein, as well as a corresponding apparatus. the method comprises the following steps: transporting a plurality of bird eggs (14), sequentially or in parallel, to an NMR apparatus (18), subjecting the bird eggs (14) to an NMR measurement, to thereby determine, for each of said eggs (14), one or more NMR parameters associated with the egg (14) selected from the group consisting of a relaxation time t1, a relaxation time t2 and a diffusion coefficient, forward said one or more nmr parameters, or parameters derived therefrom, to a classification module (38), said classification module (38) configured to determine, based on said one or more nmr parameters or parameters derived therefrom , predicting the sex of the associated egg embryo (14), and transporting said plurality of bird eggs (14) out of said NMR apparatus (18) and sorting the eggs (14) in accordance with the sex prediction. provided by said classification module (38).
公开号:BR112020008747B1
申请号:R112020008747-5
申请日:2018-11-13
公开日:2021-07-27
发明作者:Axel Haase；Benjamin Michael Schusser；Miguel Molina-Romero；Pedro A. Goméz；Maximilian Aigner；Stephan Huber；Alexander Joos
申请人:Technische Universitãt München；
IPC主号:

专利说明:

FIELD OF THE INVENTION
[1] The present invention relates to the field of aviculture for the production of laying hens and poultry. More particularly, in one aspect, the present invention relates to a method and apparatus for non-invasively determining the sex of an embryo of a bird egg, in particular a chicken egg. In one aspect, the present invention relates to a method and apparatus for non-invasively determining the fertility of a bird egg, in particular a chicken egg. BACKGROUND OF THE INVENTION
[2] In 2015, approximately 1,338 billion chicken eggs were produced. Hens raised for egg production are called laying hens. Some chicken breeds can produce more than 300 eggs per year. In rearing laying hens, right after hatching, the sex of the chicken is determined and only the hens are reared. Although roosters can also, in principle, be bred and used for meat production, specific egg laying breeds are significantly inferior for this purpose compared to specific meat production breeds, whereas specific production breeds of meat provide greater daily weight gain, greater final weight and a more favorable distribution of meat in the body. This means that, currently, when raising laying hens, males are slaughtered soon after hatching. This procedure is considered morally inappropriate by many and is at odds with current legislation and, probably, even more so with future animal protection legislation. In particular, in most countries, including Germany, it is prohibited by law to cause pain, suffering or harm to animals without just cause, and there is an ongoing debate about the extent to which this is violated in relation to the slaughter of cocks after hatching.
[3] To alleviate this problem, some attempts have been made to determine the sex of a chicken prior to hatching. For example, using molecular biological methods, it is possible to perform a PCR specifically for the female W chromosome and, based on an amplicon specific to the W chromosome, it is possible to identify female animals. In principle, it is possible to carry out these molecular biological methods even in ovo, but this would involve taking a tissue sample from the embryo and, in any case, the shell would need to be damaged.
[4] In addition to the molecular biological determination of sex in ovo, some attempts have been made to determine sex based on the content of sex hormones in the allantoic fluid of embryos, as demonstrated in Weissmann, A., Reitemeier, S., Hahn, A. ., Gottschalk, J. & Einspanier, A. Sexing domestic chicken before hatch: A new method for new gender identification, and in Theriogenology 80, pages 199 to 205 (2013), Tran, HT, Ferrell, W. & Butt, TR An estrogen sensor for poultry sex sorting. J. Anim. Sci. 88, pages 1358 to 1364 (2010). Based on a difference in the concentration of estrone sulphate, it was found that it is possible to distinguish between male and female broiler embryos. However, this method involves several problems. For example, it has been found that the hatch rate decreases due to the collection of allantoic fluid samples, probably because of the hole that needs to be formed in the eggshell. Furthermore, with this method, sex can only be determined between the 4th and 10th day of the embryo.
[5] A non-invasive possibility to determine the sex of chickens in ovo is based on the specific coloration of the feathers, according to sex. With the use of hyperspectral methods, it is possible to determine the color of the feathers through the eggshell and thus determine the sex, as demonstrated in Göhler, D., Fischer, B. & Meissner, S. In-ovo sexing of 14 -day-old chicken embryos by pattern analysis in hyperspectral images (VIS/NIR spectra): A nondestructive method for layer lines with gender specific down feather color. Poult. Sci. 96, pages 1 to 4 (2017). Although this method seems very advantageous at first glance, it requires the creation of special breeds of chicken that have specific feather markers according to sex. This technique, therefore, cannot be universally applied to all chicken breeds.
[6] An additional optical method for use in in ovo sex determination is based on Raman spectroscopy. This method has already been successfully applied to post-hatch chickens where sex was determined based on the feather follicles, see Harz, M. et al. Minimal Invasive Gender Determination of Birds by Means of UV- Resonance Raman Spectroscopy. Anal. Chem. 80, pages 1080 to 1085 (2008). For this spectroscopic method, light in the UV range was used, which unfortunately presents the risk of phototoxic effects. In Galli, R. et al. In egg sexing of domestic chicken eggs by Raman spectroscopy, Anal. Chem. 88, pages 8657 to 8663 (2016), UV radiation was used to determine sex in ovo. For this purpose, the eggshell was opened using a CO2 laser that provides access to the chicken embryo on the third and fifth days. The measurement was taken directly on an embryo's blood vessel. It was found that the mean differences in the signals of Raman spectroscopy, which allow the determination of sex, were greater in male embryos than in female embryos. The hatch rate of broilers subjected to this measurement, based on Raman spectroscopy, was only slightly lower than normal, and no negative influence was observed on the broiler's later development.
[7] In Galli, R. et al. In egg sexing of chicken eggs by fluorescence spectroscopy. Anal. Bioanal. Chem. 409, pages 1185 to 1194 (2017), fluorescence spectroscopy is used to determine the sex of chicken embryos. In this case, it was also found that it is possible to determine the sex of the embryo, on the third and fifth days, based on a pronounced fluorescence signal at 910 nm for male embryos. By combining fluorescence and Raman spectroscopy data, sex can be correctly determined in 90% of cases. However, this method still requires opening the eggshell to access the embryo's blood vessels. This presents a risk of contamination and a lower hatch rate.
[8] Document No. DE 10 2013 205 426 A1 discloses a method for determining the sex of an embryo in an egg in a non-invasive way, determining the concentration of estradiol and, optionally, in addition, the concentration of testosterone . In one embodiment, the concentration of estradiol is determined using NMR spectroscopy, based on a chemical shift. Estradiol concentration is determined before the 25th day of breeding and preferably on the 17th day of breeding. However, it was found that this method is not the best way for a practical application. One difficulty with this method is that hormone concentrations are typically only in a pMol/L range and therefore cannot be reliably detected with NMR spectroscopy.
[9] Document No. WO 00/01302 describes a non-invasive method and apparatus for sexing chicken inside the egg. The method employs high-resolution NMR imaging to determine whether the embryo within the egg contains male or female sex organs. In fact, for this method, an imaging with a spatial resolution greater than 0.1 mm is required to determine the male and female sex organs. This high resolution does not support the short measurement times needed for practical applications. In addition, high-resolution images in advanced incubation states are very sensitive to motion and therefore distorted.
[10] Consequently, neither of these two methods based on NMR spectroscopy or on NMR imaging has been found to be practically viable.
[11] In addition, in the poultry industry, it is also desirable to determine the fertility of an egg non-invasively. The poultry industry is one of the most important sources of animal protein for human consumption. The 2016 Poultry Trends magazine estimates that worldwide production and consumption of poultry meat will increase by 20% by 2025 to more than 130 million tons. In 2016, the global market produced 116.4 million tonnes of poultry meat, with the top 185 poultry producers slaughtering nearly 38 billion head to satisfy global demand. In the United States alone, the poultry industry was valued at $38.7 billion.
[12] Despite the significant volume, the process of hatching eggs to hatch poultry is far from perfect. An average poultry facility hatches only 75 to 85% of the eggs it hatches. The other 15 to 25% of eggs suffer early embryonic death or are not fertile. Currently, infertile and dead embryos are separated from live embryos after 18 days of incubation with non-invasive technologies, such as Embrex®'s egg remover (http://embrexbiodevices.com/Poultry-BioDevices/Embrex-Egg-Remove/ ). Although this solution prevents unnecessary opening of eggs, it is still a waste: all eggs without a chick inside are discarded. In other words, the industry annually incubates more than 12.8 billion eggs just to discard them. Therefore, a solution that could determine the fertility status of an egg before hatching would be highly desirable. This solution would dramatically increase productivity and save energy, costs and waste. It would also incorporate billions of eggs on the market for human consumption.
[13] There are several patents in the field of identifying fertile eggs. For example, US Patent No. 5,745,228 - Method and apparatus for distinguishing live from infertile poultry eggs uses a light source to determine whether the poultry within the egg is alive. This is the technology used in Embrex®'s egg remover.
[14] Patent No. US 6,029,080 - Method and apparatus for avian pre-hatch sex determination proposes the use of RM for sexing members of avian species from an egg. Although this patent uses MR technology, it does not refer to the determination of fertility status and exclusively addresses the identification of the gonads through MR for sexing.
[15] Patent No. US 7,950,439 B1-Avian egg fertility and gender detection suggests the use of an external light source in the form of incandescent, fluorescent or LED lights for the determination of fertility and gender of an avian egg .
[16] Patent No. US 6,535,277 B2- Methods and apparatus for non-invasively identifying conditions of eggs via multi-wavelength spectral comparison is based on the use of visible and invisible light at wavelengths between 300 nm and 1,100 nm to identify the multiple conditions of an egg, which includes the fertility status.
[17] US Patent No. US 2013/0044210 AI - Hyperspectral identification of egg fertility and gender uses mid-IV light to determine the fertility of an egg. The inventors of this patent claim that they are able to determine the fertility status of an egg on day zero (ie, newly laid) with an accuracy of 90%. SUMMARY OF THE INVENTION
[18] The problem underlying a first aspect of the invention is to provide a method and apparatus for automated, non-invasive determination of the sex of an embryo from a bird egg, in particular a chicken egg, which allows for rapid determination and reliable sex of the embryo at an early stage, in which the embryo has not yet developed a feeling of pain. This problem is solved by a method according to claim 1 and an apparatus according to claim 25. Preferred embodiments are defined in the dependent claims.
[19] The method, according to the first aspect of the invention, comprises the steps of transporting a plurality of avian eggs, sequentially or in parallel, to an NMR apparatus and subjecting the avian eggs to an NMR measurement, and thus, determining, for each of said eggs, one or more NMR parameters associated with the egg selected from the group, which consists of a relaxation time T1, a relaxation time T2 and a diffusion coefficient.
[20] In the present document, each of these "parameters" can correspond to a parameter value representative for a region of interest within the egg, such as the exact region or close to the region where the embryo is located. For example, the “T1 parameter” can refer to an average value of the measured T1 values in the region of interest.
[21] However, as used in this document, the term "parameter" can also refer to a parameter image of a region of said egg, where the parameter values are associated with the corresponding pixels or voxels in the image. Note that the term "image" does not imply that it is something that must necessarily be visually inspected, but merely implies that a parameter value is associated with a particular spatial region within the egg, where the region is usually called, in an image, of a pixel or voxel. Since in different embodiments of the invention, for each parameter (such as T1, T2 or diffusion coefficient), a parameter value, or a parameter image, can be employed, both variants are generally referred to herein as " parameter", for the sake of simplicity.
[22] The method further comprises forwarding said one or more NMR parameters, or parameters derived therefrom, to a classification module, said configuration module being configured to determine, based on said one, or more NMR parameters or parameters derived therefrom, predicting the sex of the associated egg embryo and transporting said plurality of avian eggs out of said NMR apparatus and sorting the eggs, according to the sex prediction provided by said classification module. In this document, the expression "or parameters derived from them" indicates that, instead of the parameter value itself, a value derived from it can be used for classification, for example, a normalized value, the square of value or similar, which may be more appropriate for classification purposes.
[23] According to the present invention, sex determination, through the classification module, is performed based on one or more specific NMR parameters, namely, a relaxation time T1, a relaxation time T2 and a diffusion coefficient. Surprisingly, each of these NMR parameters was found to be sensitive to the sex of a chicken embryo. Consequently, based on at least one, but typically a set of more than one of these parameters, the sex of the embryo within the egg can be determined using a classification module.
[24] Note that this method is different from prior art methods, based on determination of estradiol concentration using NMR spectroscopy, or based on sex determination, analyzing high-resolution NMR images to recognize sufficiently developed sex organs, which can be considered as "deductive" methods, in the sense that certain sex-related characteristics (hormones, sex organs) are present and are then verified using NMR techniques. In comparison to that, the method of the invention is based only on parameters, without the need for any theory or explanation as to why these parameters are correlated with the sex of the embryo. Rather, the method of the invention is based on the surprising observation that the three NMR parameters mentioned above are characteristic for the sex of the chicken embryo, even at a very early stage. According to current understanding, the best prediction can be made if the rating unit receives a set that comprises all three of the parameters mentioned above, T1, T2 and diffusion coefficient, and bases the prediction on them, for example, with the use of a suitable classifier. However, it is not necessary to combine all three of these parameters into one set, instead you can use just the combination of two of these parameters or one of these parameters in combination with additional NMR parameters. In fact, specifically the T1 parameter of NMR is a sufficient sexual characteristic for a prediction, based only on this parameter, to be conceivable.
[25] The advantage of having a classification module based on decision parameters allows a good fit between high throughput and consistent and reliable classification. In general, the measurement and processing of some individual parameters can be carried out comparatively quick and easy, for example, compared to the high resolution imaging that needs to be employed in document No. WO 00/01302, as a consistent classification can be supported, particularly, through gender differences in T1, in T2 and in the diffusion coefficient. In fact, the various modalities described below make it possible to determine the sex of each of, for example, 150 eggs, in parallel, in less than three minutes, with the potential to reduce this measurement time to two minutes and possibly even to one minute when optimal use is made of rapid MR techniques and parallel imaging, which thus allow an average measurement time of one second per egg or less. In addition, confidence can be increased by increasing the number of parameters in the parameter set, which can only equate to a moderate increase in measurement time, and still significantly increase the predictive quality of the method, even if the inter- relation of parameters and its relation with sex are not fully understood. This is fundamentally different when compared to purely deductive methods, where the conclusion about sex must be reached on certain biological premises (presence of estradiol, absence of testosterone, presence of sexual organs in the image). In such deductive methods, confidence can only be increased by increasing confidence in the given assumptions, where a small degree of additional confidence can lead to extraordinary additional NMR measurement efforts and, consequently, reduced method throughput. Furthermore, other indications beyond the premises remain without consideration, in particular indications that could be developed at an early stage of the embryo, other than hormonal concentration or sexual organs. For example, according to document No. DE 10 2013 205 426 A1, the concentration of estradiol is preferably determined on the 17th day of reproduction, when the chicken embryo already feels pain.
[26] In a preferred embodiment, said one or more NMR parameters comprises a set of two or more NMR parameters, at least one being selected from said group, consisting of a relaxation time T1, a relaxation time T2 and a diffusion coefficient.
[27] In a preferred embodiment, said set of NMR parameters additionally comprises one or more of the following parameters: a relaxation time T2*, a relaxation time Tip and a spin density associated with one or more of the nuclei 1H, 13C, 23Na and 31P, or parameters derived therefrom. Although these parameters, according to the inventors' current understanding, are less characteristic for the sex of the embryo in the early stage of development, however, they can be combined with some or all of the three favorite parameters T1, T2 and diffusion coefficient , in a set, and can be considered through the classification module, thus increasing the reliability of the determination.
[28] Additionally or alternatively, said set of NMR parameters preferably further comprises one or more chemical shift signals from metabolites, in particular water, lipids, amino acids, nucleic acids or hormones; a chemical shift selective transfer signal; and NMR signals of zero quantum coherence or multiple quantum coherence or parameters derived therefrom. For example, in case of a chemical shift signal spectrum, a “derived parameter therefrom” can be the amplitude of a given peak, the ratio of the two peaks, the difference of the two peaks, or the like. A parameter can be a number or it can be a set of numbers, such as a vector. However, if the classification module is a machine learning module, as explained below, it is also possible to simply provide the entire spectrum to the module, which through machine learning, is able to determine the relevant characteristics on its own. the same.
[29] In a preferred modality, said classification module is a machine learning module. The inventors have found machine learning to be a particularly efficient way to determine sex based on one or more of the parameters T1, T2 and diffusion coefficient mentioned above. Namely, although the available measures clearly indicate that these parameters are characteristic for the sex of the embryo, there is currently no deductive biological model available on the exact reason or by what underlying biological mechanism sex is related to each of these parameters, or to the distribution. of the parameters inside the egg, as represented in a parameter image. Consequently, machine learning is ideal for recognizing patterns in any of these parameters or combinations thereof that are indicative of the sex of the embryo that allow for reliable predictions and early prediction dates in the embryo's life. If the classification module is a machine learning module, said NMR parameter values, or parameters derived therefrom, can form feature values presented to the machine learning module as a feature vector. In other words, a representative parameter value for a region of interest within the egg or a parameter image of an egg region can be presented to the machine learning module as a feature vector or as part of a feature vector.
[30] In a preferred embodiment, the classification module is configured to determine the prediction of embryo sex using a linear classifier. Preferred linear classifiers, for purposes of the present invention, are based on one or more linear regressions of least squares, nearest neighbors, logistic regression, separation hyperplanes, or perceptrons.
[31] In an alternative preferred modality, said classification module is configured to determine the prediction of embryo sex using a non-linear classifier. Preferred nonlinear classifiers, for the purposes of the present invention, are based on per part polynomials, splines, kernel smoothing, tree-based method, support vector machines, random forest, reinforcement, additive and set methods or graphics templates.
[32] In particularly advantageous modalities, the classification module is configured to determine the prediction of the embryo's sex using a deep learning algorithm, which allows to optimize the use of information in relation to sex, as expressed in the parameter set. of NMR. Preferred deep learning algorithms for the purpose of the invention are based on convolutional neural networks, recurrent neural networks, or long-short-term memory networks.
[33] In alternative embodiments, the classification module can be configured to determine the prediction of embryo sex based on a comparison with parameter values stored in a database.
[34] In preferred embodiments, measurement of NMR comprises NMR imaging, in which an NMR imaging plane is arranged to cross the location of the embryo. In the present document, NMR imaging can in particular refer to imaging parameters for one or more of the parameters T1, T2 and diffusion coefficient.
[35] In a preferred embodiment, eggs are laid out in a regular pattern, in particular in a matrix configuration, on a tray during said transport and NMR measurement.
[36] Preferably, the number of eggs arranged in said tray is at least 36, preferably at least 50 and more preferably at least 120.
[37] In a preferred modality, the method is performed before the eighth day of reproduction, preferably on the fifth day of reproduction, when the embryo has not yet developed any pain sensation.
[38] The problem underlying a second aspect of the invention is to provide a non-invasive technique that is able to automatically identify infertile eggs, immediately after laying and before incubation, and is able to handle a high yield of eggs and not damage or change the eggs in any way. This objective is achieved by means of a method, according to claim 7, and an apparatus, according to claim 15. Preferred embodiments are defined in the dependent claims.
[39] The method according to the second aspect of the invention comprises the steps of
[40] transport a plurality of bird eggs, sequentially or in parallel, in an NMR machine,
[41] subjecting bird eggs to an NMR measurement to thus determine for each of said eggs, one or both:
[42] - a histogram of diffusion coefficients at various locations in the egg,
[43] - an NMR spectrum of the yolk, which includes peaks, which correspond to water and fat,
[44] Determine fertility prediction based on scattering coefficient histogram format and/or NMR spectrum, and
[45] transporting said plurality of bird eggs (14) out of said NMR apparatus (18) and sorting the eggs in accordance with the fertility prediction.
[46] The inventors have found that, surprisingly, the shape of a diffusion coefficient histogram differs for fertile and infertile eggs at various locations in the egg. In this document, the histogram indicates how often certain diffusion coefficients occur when measurements are taken at various locations in the egg. Consequently, by analyzing the format of the diffusion coefficient histogram, fertility can be predicted.
[47] Furthermore, the inventors have found that equally, surprisingly, the NMR spectra of fertile and infertile egg yolks differ with respect to peaks corresponding to water and fat. Consequently, the shape of an NMR spectrum, which includes such water and fat peaks, are similar characteristics of fertility and can be used in the determination.
[48] Although only one of the two fertility traits can be used in the method, in preferred modalities both traits are combined in the determination, thus increasing the reliability of the prediction.
[49] Since eggs can be subjected to NMR measurements without causing any harm or damage to the shell or interior, the hatch rate is not adversely affected by this measurement. At the same time, unnecessary incubation of infertile eggs can be avoided. Furthermore, since infertility is determined before incubation, it has been found that these infertile eggs can still be used for feeding, which is not possible after the start of incubation.
[50] Although there are many ways to analyze the shape of a diffusion coefficient histogram, in a preferred modality, determining fertility, based on the format of the diffusion coefficient histogram, comprises comparing the frequency of occurrence of at least two different diffusion coefficients or ranges of diffusion coefficients. This is a particularly simple way of characterizing the shape of the diffusion coefficient histogram, which has proven to give surprisingly reliable results.
[51] In a preferred embodiment, said at least two different diffusion coefficients or the centers of said at least two diffusion coefficient ranges are separated between 0.5 and 2.5 mm2/s, preferably between 0.75 and 1.5 mm2/s.
[52] Of said at least two different diffusion coefficients, or the centers of said at least two diffusion coefficient ranges, preferably one is located in a range of 0.6 to 1.3 mm2 s, preferably in a range from 0.7 to 1.2 mm2/s, and the other is located in a range of 1.5 to 2.5 mm2 s, preferably in a range of 1.7 to 2.3 mm2/s.
[53] In a preferred embodiment, said various locations in the egg are evenly distributed in the egg and, in particular, correspond to voxels of a diffusion coefficient image.
[54] Regarding spectrum, the inventors have observed that when the spectrum is, for example, normalized to peaks corresponding to fat, the peak corresponding to water is larger in an infertile egg compared to a fertile egg. Consequently, one way to determine fertility is through the ratio of peak water to fat. However, there are different ways to classify fertility based on NMR spectra. In particular, it is possible to present the spectrum, or certain characteristics of the spectrum, such as peak heights and peak locations, to a machine learning module that performs the classification.
[55] In either method, according to the first or second aspect of the invention, the eggs are arranged in a regular pattern, in particular in a matrix configuration, on a tray during said transport and NMR measurement . Preferably, the number of eggs arranged in said tray is at least 36, preferably at least 50 and more preferably at least 120.
[56] The invention further relates to an apparatus for automated, non-invasive sex determination of a bird egg embryo, comprising: an NMR apparatus and a transport device for transporting a plurality of avian eggs, sequentially or in parallel, into and out of said NMR apparatus. The NMR apparatus is configured to subject the avian eggs to an NMR measurement to thereby determine, for each of said eggs, one or more NMR parameters associated with the selected egg from the group consisting of a time of relaxation T1, a relaxation time T2 and a diffusion coefficient, where each of said parameters corresponds to:
[57] - a representative parameter value for a region of interest within the egg, or
[58] - a parameter image of a region of said egg, in which the parameter values are associated with the corresponding pixel or voxel of the image.
[59] The apparatus further comprises a classification module configured to receive said one or more NMR parameters or parameters derived therefrom, wherein said classification module configured to determine the associated egg embryo sex prediction based on said one or more NMR parameters or parameters derived therefrom. Finally, the apparatus comprises an egg separating device for separating eggs in accordance with the sex prediction provided by said classification module.
[60] In a preferred embodiment, said one or more NMR parameters comprises a set of two or more NMR parameters, at least one being selected from said group consisting of a relaxation time T1, a time of T2 relaxation and a diffusion coefficient.
[61] Preferably, said set of NMR parameters further comprises one or more of the following parameters: a relaxation time T2*, a relaxation time Tip and a spin density associated with one or more of the 1H,13C nuclei , 23Na and 31P.
[62] Additionally or alternatively, said set of NMR parameters preferably further comprises one or more chemical shift signals from metabolites, in particular water, lipids, amino acids, nucleic acids or hormones; a chemical shift selective transfer signal; and NMR signals of zero quantum coherence or multiple quantum coherence.
[63] In a preferred modality, said classification module is a machine learning module.
[64] Preferably, said NMR parameter values or parameters derived therefrom form feature values presented to the machine learning module as a feature vector.
[65] In a preferred embodiment, said classification module is configured to determine the prediction of embryo sex using a linear classifier, in particular, a linear classifier based on one or more neighboring least square linear regressions nearest, logistic regression, separation hyperplanes or perceptrons.
[66] In a preferred alternative modality, said classification module is configured to determine the prediction of embryo sex using a non-linear classifier, in particular a non-linear classifier based on polynomials per part, splines, smoothing by kernel, tree-based method, support vector machines, random forest, reinforcement, additive and sets methods or graphics models.
[67] Still in a preferred alternative modality, said classification module (38) is configured to determine the prediction of the embryo's sex using a deep learning algorithm, in particular, a deep learning algorithm based on networks convolutional neural networks, recurrent neural networks or long-short-term memory networks.
[68] In an alternative embodiment, the classification module (38) is configured to determine the prediction of the sex of the embryo based on a comparison with parameter values stored in a database.
[69] The invention further relates to an apparatus for automated, non-invasive determination of the fertility of a bird egg, comprising an NMR apparatus and a transport device for transporting a plurality of bird eggs sequentially or in parallel, into and out of said NMR apparatus, wherein said NMR apparatus is configured to subject avian eggs to an NMR measurement to thereby determine, for each of said eggs, one or both:
[70] - a histogram of diffusion coefficients at various locations in the egg,
[71] - an NMR spectrum of the yolk, which includes peaks, which correspond to water and fat,
[72] wherein said apparatus is further configured to determine a fertility prediction based on the scatter coefficients histogram format and/or NMR spectrum, wherein said apparatus further comprises an egg separating device for separating the eggs, according to the fertility prediction.
[73] In a preferred embodiment, said fertility determination, based on the diffusion coefficient histogram format through the device, comprises comparing the frequency of occurrence of at least two different diffusion coefficients or diffusion coefficient ranges.
[74] In a preferred embodiment, said at least two different diffusion coefficients or the centers of said at least two diffusion coefficient ranges are separated between 0.5 and 2.5 mm2/s, preferably between 0.75 and 1.5 mm2/s.
[75] In a preferred embodiment, of said at least two different diffusion coefficients, or the centers of said at least two diffusion coefficient ranges, one is located in a range of 0.6 to 1.3 mm2 s, preferably in a range of 0.7 to 1.2 mm2/s, and the other is located in a range of 1.5 to 2.5 mm2 s, preferably in a range of 1.7 to 2.3 mm2/s.
[76] In a preferred embodiment, said various locations in the egg are evenly distributed in the egg and, in particular, correspond to voxels of a diffusion coefficient image.
[77] In a preferred embodiment, said apparatus is configured to determine a fertility prediction based on the NMR spectrum, based on the ratio between the peaks corresponding to water and fat in said NMR spectrum.
[78] Regardless of whether the apparatus is configured to determine the sex of the embryo or the fertility of the egg, in preferred embodiments, the apparatus additionally comprises a tray, on which said eggs can be arranged in a regular pattern, in particular, in a matrix configuration, during said transport and NMR measurement.
[79] In a preferred embodiment, the number of eggs that can be disposed on said tray is at least 36, preferably at least 50 and more preferably at least 120.
[80] In a preferred embodiment, said NMR apparatus comprises an array of RF coils to apply RF magnetic fields to eggs located in the tray and/or to detect NMR signals, said array of RF coils comprising one or more:
[81] - a plurality of coils arranged in a plane, located above the tray, loaded with eggs, when transported to the NMR apparatus,
[82] - a plurality of coils arranged in a plane, located under the tray, loaded with eggs, when transported to the NMR machine,
[83] - a plurality of coils arranged in vertical planes, which extend between rows of eggs in the tray, when transported to the NMR apparatus, where the rows extend parallel to the direction of transport of the tray into and out of the NMR machine.
[84] In case of a plurality of coils arranged in a plane located above or below the tray loaded with eggs, the ratio between the number of coils and the number of eggs arranged in said tray is between 1:1 and 1:25, preferably between 1:1 and 1:16, and more preferably between 1:1 and 1:5.
[85] In a preferred embodiment, said NMR apparatus comprises an array of RF coils to apply RF magnetic fields to eggs located in the tray (16) and/or to detect NMR signals, said coil array of RF is integrated or attached to said tray.
[86] In the present document, the tray preferably comprises a plurality of cavities or pockets for receiving the corresponding egg, wherein a number of spools is associated with each of said cavities or pockets, wherein said number of spools per cavity or pocket is at least 2, preferably at least 3 and more preferably at least 4 and/or wherein at least some of said spools are disposed vertically with respect to the main plane of the tray, or at an angle of at least 50°, preferably at least 75° and more preferably at least 80° with respect to the main plane of the tray. BRIEF DESCRIPTION OF THE DRAWINGS
[87] Figure 1 is a schematic illustration of an apparatus for automated, non-invasive determination of the sex of an embryo from a bird egg or the fertility of a bird egg.
[88] Figure 2A is a perspective view of an array of RF coils arranged in a parallel plane and just above a tray loaded with eggs.
[89] Figure 2B is a perspective view of an array of RF coils arranged in a parallel plane and just above a tray loaded with eggs.
[90] Figure 2C is a perspective view of an array of RF coils, where the coils are arranged in vertical planes that extend between rows of eggs on the tray, in which the rows extend parallel to the direction of transport from the tray.
[91] Figure 3 schematically shows a portion of a tray that includes a cavity for receiving an egg and four RF coils integrated into the tray surrounding the egg.
[92] Figure 4 additionally shows the details of the coil arrangements of Figures 2A to C and 3.
[93] Figure 5 shows six images of NMR parameters taken by the NMR machine of Figure 1.
[94] Figure 6 shows pairwise combinations of the parameters T1, T2 and D in the histograms and off-diagonal diagrams for each of the parameters T1, T2 and D in the diagrams arranged along the diagonal.
[95] Figure 7 shows a histogram averaged for T1 values for male and female embryos.
[96] Figure 8 shows a histogram averaged for T2 values for male and female embryos.
[97] Figure 9 shows a histogram averaged for the diffusion coefficients for male and female embryos.
[98] Figure 10 shows an averaged histogram of the observed diffusion coefficient for a plurality of fertile eggs (solid line) and infertile eggs (dotted line).
[99] Figure 11 is a scatter plot showing pairs of diffusion coefficient histogram values at 1 mm/s and 2 mm/s for a plurality of eggs.
[100] Figure 12 shows NMR spectra for fertile and infertile eggs.
[101] Figure 13 is a schematic illustration of a simplified apparatus for non-automated, non-invasive determination of the sex of an embryo from a bird egg or the fertility of a bird egg.
[102] Figure 14 is a schematic illustration of an architecture of a machine learning classifier for egg sex determination based on convolutional neural networks. DESCRIPTION OF THE PREFERRED MODALITY
[103] In order to promote an understanding of the principles of the invention, reference will now be made to a preferred embodiment illustrated in the drawings, and specific language will be used to describe the same. However, it will be understood that no limitation on the scope of the invention is thus intended, and such additional changes and modifications to the illustrated apparatus and such additional applications of the principles of the invention, as illustrated herein, are contemplated as would normally now occur or in the future to a person skilled in the art to which the invention relates.
[104] Figure 1 shows a schematic representation of an apparatus 10, according to a preferred embodiment of the invention. Apparatus 10 comprises a transport device 12 for transporting a plurality of eggs 14 arranged in a matrix configuration on a tray 16, into and out of an NMR apparatus 18, which is represented by the hatched box in the Figure. In the embodiment shown, the transport device 12 comprises a conveyor belt 20, on which trays 16 can be transported. The movement of the conveyor belt 20 is controlled by a corresponding conveyor controller 22.
[105] The NMR apparatus 18 comprises a magnet arrangement 24 to provide an external magnetic field in the z direction with which nuclear spins can interact. The z direction of the magnetic field coincides with the direction of transport on the conveyor belt 20, but this is not crucial to the function of the apparatus 10. In the embodiment shown, the magnet arrangement 22 generates a static magnetic field having a field strength of 1T, but the invention is not limited to that. Instead, a wide range of magnetic field strengths can be used and, in alternative embodiments of the invention, even the strength of the earth's magnetic field may suffice, as shown in Stepisnik, J., Erzen, V. & Kos , M. NMR imaging in the earth's magnetic field. Magn. Reson. Med.15, pages 386 to 391 (1990), and Robinson, J.N. et al. Two-dimensional NMR spectroscopy in Earth's magnetic field. J. Magn. Reson.182, pages 343 to 347 (2006).
[106] Furthermore, the NMR apparatus 18 comprises gradient coils 26 which are used to generate spatial gradient fields which are used for image encoding or, in other words, NMR measurements with spatial resolution, in a per se manner. known to the person versed, and described in Lauterbur, Image formation by induced local interactions. Examples using nuclear magnetic resonance. Nature 242, pages 190 to 191 (1973). In addition, the gradient coils 26 are also used to increase the local homogeneity of the external magnetic field created by the magnetic array 24. The gradient fields applied by the gradient coils 26 are controlled by a gradient controller 28. In the embodiment shown, the gradient controller 28 is optimized for efficient coverage of the measurement space (the k space) in order to increase measurement speed. In particular, gradient controller 28 is preferably configured to perform ecoplane imaging. For details of ecoplanar imaging, reference is made to Stehling, M., Turner, R. & Mansfield, P. Echo-planar imaging: magnetic resonance imaging in a fraction of a second. Science (80-.).254, pages 43 to 50 (1991), and Mansfield, P. & Maudsley, AA Planar spin imaging by NMR. J. Phys. C Solid State Phys.9, pages L409 to L412 (1976). Alternatively, the gradient controller 28 can control the gradient coils 26 to perform spiral readings with ideal gradient design over time, as described in Hargreaves, BA, Nishimura, DG & Conolly, SM Time-optimal multidimensional gradient waveform design for rapid imaging. Magn. Reson. Med. 51, pages 81 to 92 (2004), which allows very fast NMR imaging.
[107] Multiple RF coils 30 are arranged to surround the tray 16 loaded with eggs 14 on the conveyor belt 20 when the tray 16 is transported to the NMR apparatus 18. As the skilled person will note, the RF coils 30 are used to provide RF pulses that excite spins and, in particular, the spins of hydrogen atoms inside eggs 14. The timing, shape and intensity of the pulses are controlled by the RF controller 32. A serial manipulation of the RF pulses and gradients allow modulation of the measured signal for fast image encoding. In order to allow for high throughput measurements, rapid pulse sequences such as small-angle rapid firing imaging or quantitative transient imaging can be deployed, as described in more detail in the articles Haase, A., Frahm, J., Matthaei, D., Hanicke, W. & Merboldt, FLASH imaging. Rapid NMR imaging using low flip-angle pulses. J. Magn. Reson.67, pages 258-266 (1986) and Gómez, PA et al. Accelerated parameter mapping with compressed sensing: an alternative to MR Fingerprinting. Proc Intl Soc Mag Reson Med (2017), co-authored with the present inventors and included in this document, by way of reference. These fast pulse sequences are designed to be sensitive to different relevant parameters employed in the present invention, in particular T1 and T2 relaxation and diffusion, but also to water and fat content or to magnetization transfer.
[108] In addition, the precession movement of the spins excited in the external magnetic field, provided by the magnetic array 24, leads to current flow in the RF coils 30, which can be detected by an RF detector or 34. RF 34 translates the current flow of the RF 30 coils into an interpretable signal. This includes analog to digital conversion, demodulation and signal amplification.
[109] The NMR device 18 additionally comprises an image reconstruction module 36. In preferred embodiments, measurements from different RF coils 30 will be combined using parallel imaging techniques, and an image reconstruction is achieved through the application Fast Fourier Transform (FFT) on the acquired measurements. For details of parallel imaging techniques, reference is made to Pruessmann, KP, Weiger, M., Scheidegger, MB and Boesiger, P. SENSE: sensitivity encoding for fast MRI. Magn. Reson. Med.42, pages 952 to 962 (1999), and Uecker, M. et al. ESPIRiT - An eigenvalue approach to autocalibrating parallel MRI: Where SENSE meets GRAPPA. Magn. Reson. Med.71, pages 990 to 1001 (2014).
[110] When non-Cartesian sampling is employed, non-uniform FFT, as described in Fessler, JA and Sutton, B. Nonuniform Fast Fourier Transforms Using Min-Max Interpolation. IEEE Trans. Signal Process. 51, pages 560 to 574 (2003) can be employed. In the embodiment shown, the image reconstruction module 36 implements advanced reconstruction algorithms such as reduced rank matrix recovery or iterative algorithms. The image reconstruction module 36 is configured to process data of different dimensions, namely, 1D or 2D NMR signals, 2D images, 3D volumes and 4D time series.
[111] The data processed by the image reconstruction module 36 is transmitted to an egg classification module 38. In the embodiment shown, the egg classification module 38 has two purposes, segmentation and classification. In the high-throughput device, the egg classification module 38 first segments the received images into image portions corresponding to individual eggs 14. Then, the image portion corresponding to each individual egg 14 is classified according to its sex of a way to be described in more detail below.
[112] The egg sorting result is fed to an egg sorting device 40. In the embodiment shown, the sorting result is fed to the egg sorting device 40 in the form of a matrix containing the coded sexes of the eggs 14 in a given tray 16. Based on this information, the egg separating device 40 can separate the eggs 14 determined, such as including male embryos, or can rearrange the eggs 14 in the tray 16 according to sex. As shown schematically in Figure 1, the egg separator 40 has as many containers 42 as eggs 14 in tray 16, wherein said containers 42 are connected to a vacuum device (not shown). When a container 42 is moved close to the corresponding egg 14, the egg 14 will be attracted and fixed to the container 42 by vacuum suction so that it can be safely collected and carefully placed in a different location.
[113] Finally, a central controller 44 is provided, which is connected for data communication with each of the above mentioned components involved in the measurement by NMR, egg classification by image reconstruction and egg separation process, through the channels of corresponding data 45.
[114] The 18 NMR device developed for the classification of eggs in an industrial environment deals with a well-defined scanning geometry. Eggs 14 are introduced into the NMR apparatus 18 arranged in a matrix configuration with M rows and N columns in a corresponding tray 16, where the columns are arranged parallel to the direction of transport on the conveyor belt 20 of Figure 1. Various embodiments of invention employ an array 30 of RF coils 30a that is designed to maximize the signal to noise ratio and minimize the acquisition time, which will be described below with reference to Figures 2 to 4. radio frequency decays with the square of the distance from the emitting source, the preferred designs aim to place the RF coils 30a as close as possible to the eggs 14. In addition, having an array 30 of coils 30a creates spatial redundancy in the receiving field that can be exploited to reduce scanning time.
[115] Figures 2A to 2C show three different RF coil arrangements 30 that are particularly suited to establishing preferred signal-to-noise ratios and minimum acquisition times. In each of Figures 2A to 2C, an array 30 of RF coils 30a is schematically shown together with tray 16 loaded with eggs 14. Each individual RF coil 30a of RF coil array 30 is shown as having a geometry for simplicity, but different geometries can also be implemented. In the embodiment of Figure 2A, the individual RF coils 30a are arranged in a parallel plane and slightly above tray 16. The number of individual RF coils 30a may, but need not, correspond to the number of eggs 14. Preferably, the ratio between the number of RF coils 30a and the number of eggs 14 arranged in tray 16 is between 1:1 and 1:25, preferably between 1:1 and 1:16 and most preferably between 1:1 and 1:5. Each of the individual RF coils 30a is connected via a corresponding transmission line 30b with the RF controller 32 and the RF detector 34. Although, in the simplified Figures, all transmission lines 30b are shown as a single cable. , it is to be understood that such cable includes a plurality of individual wires so that each RF coil 30a of the RF coil array 30 can be individually controlled by RF controller 32 and read by RF detector 34. Arrow 46 indicates the direction of transport direction of tray 16 by transport device 12.
[116] Figure 2B shows a similar RF coil arrangement 30 as in Figure 2A, which is, however, placed under tray 16.
[117] Figure 2C shows an array 30 of RF coils of RF coils 30a, arranged vertically and placed next to eggs 14, rather than above or below, as is the case in Figures 2A and 2B. In order not to interfere with the movement of eggs 14 on conveyor belt 20, the RF coils 30a of the RF coil arrangement 30 are arranged in vertical planes extending between the rows of eggs 14 on the tray 16, whose rows extend parallel to the direction tray 16 by transport device 12 as indicated by arrow 46.
[118] As the embryo always floats to the top of egg 14, the area of interest is mostly located in the upper half of the egg. This implies that the configurations of Figure 2A (top plane of the RF coil array 30) and Figure 2C (RF coils 30a arranged in vertical longitudinal planes) allow smaller distances between the RF coils 30a of the RF coil array 30 and regions of interest in eggs 14 than the configuration in Figure 2B and therefore a favorable signal-to-noise ratio. However, in various embodiments, the RF coil array 30 disposed in a plane below the tray 16, as shown in Figure 2B, can be used in place of or in combination with any of the configurations in Figure 2A and 2C. In fact, any two or all three configurations of Figures 2A, 2B and 2C can be combined in the 18 NMR machine.
[119] In an alternative embodiment, the RF coils 30a are connected or integrated with the tray 16, as shown in Figure 3. Figure 3 schematically shows a portion of the tray 16, in which a cavity 48 is formed to receive an egg 14 Four RF coils 30a are attached or integrated to tray 16 around the egg 14. Generally, one or more RF coils 30a per cavity 48 can be provided. Other particularly favorable arrangements provide three, five, six or eight RF coils 30a per cavity 48. The connection or integration of the RF coils 30a with the tray 16 allows a denser integration of the RF coils 30a and shorter distances with the corresponding eggs 14, without interfering with the transport of the eggs 14 in the tray 16, which allows particularly high signal-to-noise ratios and minimized acquisition time. However, in this modality, eggs 14 need to be transferred from the transport trays (not shown) to the specific NMR trays 16 and later to the setter trays (not shown).
[120] Figure 4 shows more details of the RF coil arrays 30, which can be applied independently of the particular geometric arrangement of the RF coils 30a in the RF coil array 30, and therefore can be applied to any of the modalities shown in Figure 2A, 2B, 2C and 3. As shown schematically in Figure 4, each of the RF coils 30a may comprise an antenna section 50 which, in the embodiments shown, is in the form of a circular circuit. However, antenna sections 50 with different geometries such as Helmholtz coils, solenoid coils, saddle coils or cage coils can also be employed.
[121] In addition, each RF coil 30a comprises a tuning capacitor 52 to reduce mutual inductance and tune the center frequency, and a preamplifier 54 that improves tuning, matching, and decoupling. In addition, each RF coil 30a is connected via transmission lines 56 with a multi-channel NMR spectrometer 58, which combines the functionality of the RF controller 32 and the RF detector 34 shown in Figure 1.
[122] Obviously, the NMR measurement time is critical for a high-throughput device. The preferred embodiments of the invention are therefore optimized for high-speed acquisition and reconstruction. In particular, the RF coil arrangements 30 described above are suitable for parallel imaging, to thereby acquire less information per RF coil 30a and combine it with use of spatial redundancy, such as speeding up the measurement.
[123] Preferred embodiments of the invention employ the so-called SENSE method, described in Pruessmann, KP, Weiger, M., Scheidegger, MB & Boesiger, P. SENSE: sensitivity encoding for fast MRI. Magn. Reson. Med. 42, pages 952 to 962 (1999), which uses spatial redundancy to acquire a subsample of k space and reconstruct unaliased images. A related method that is equally applicable is the so-called Generalized Autocalibrating Partially Parallel Acquisition (GRAPPA) method, as described in Griswold, MA et al., Generalized auto calibrating partially parallel acquisitions (GRAPPA), Magn. Res. Med 47, pages 1202 to 1210 (2002).
[124] To further increase throughput, multi-band technologies are employed, which employ multiple excitation frequencies to allow parallel acquisition at different spatial locations along the orifice of the magnetic array 42, thus reducing the total time of sweep. A more detailed explanation of multiband technology is given in Feinberg, D.A. et al. Multiplexed echo planar imaging for sub-second whole brain fmri and fast diffusion imaging. PLoS One 5, (2010), which is included here for reference.
[125] On top of these techniques, in the preferred modalities, the so-called compressed detection is employed, which reduces the number of measurement points needed to reconstruct an image, thus introducing a new acceleration factor. A determination of compressed sensing is given in Lustig, M., Donoho, D. & Pauly, JM Sparse MRI: The application of compressed sensing for rapid MR imaging. Magn. Reson. Med. 58, pages 1182 to 1195 (2007).
[126] In addition, in preferred modalities, imaging is performed in the transient state, which can be performed ultra-fast and using quantitative parameters, as described in the papers co-authored with the present inventors, see Gómez, PA et al. Accelerated parameter mapping with compressed sensing: an alternative to MR Fingerprinting. Proc Intl Soc Mag Reson Med (2017). Another suitable way of transient imaging is described in Ma, D. et al. Magnetic resonance fingerprinting. Nature 495, pages 187 to 192 (2013).
[127] The RF coil array configurations and image reconstruction methods introduced earlier allow for rapid imaging of the 3D space that contains the NxM 14 egg array. Depending on the geometry of the RF coil and the processing method chosen, in some modalities, one image per egg 14 will be reconstructed, while in other modalities, a single image per tray 16 will be reconstructed. In the case of a single image per egg 14, each image can be sorted individually. In the case of an image per tray 16, the individual eggs 14 in the image must first be segmented before sorting. There are several segmentation techniques that can be employed; but, given the simplicity of the geometry of the trays 16, the preferred solution is to predefine a grid corresponding to each cavity 48 with a single egg 14.
[128] Figure 5 shows six images of NMR parameters taken by the MRI machine 18 of Figure 1. The upper row of images corresponds to an egg that includes a seven-day-old female embryo, while the lower row of images corresponds to to an egg, which includes a seven-day-old male embryo. As explained in the introductory part of the application, a "parameter image" of a region of egg 14, as referred to herein, means a set of parameter values that are associated with corresponding spatial regions in egg 14, which correspond to pixels or image voxels. Each of the images shown in Figure 5 consists of 64 x 64 voxels of a thick section of a voxel through the egg, to which the value of the corresponding parameter T1 (left column), T2 (middle column) or diffusion coefficient D (right column) is associated and shown in grayscale in Figure 5.
[129] In this document, T1 usually denotes the time constant for the physical process responsible for relaxing the components of the nuclear spin magnetization vector parallel to the external magnetic field generated by magnetic array 24, which is also known as the “longitudinal relaxation time” or “spin lattice time” in the technique. It is, therefore, the time it takes for the longitudinal magnetization to recover approximately 63% (1-(1-/e)) of its initial value, after being inverted in the magnetic transverse plane by a 90° radiofrequency pulse.
[130] T2 denotes the "transverse relaxation time" or "spin-spin" and represents the decay constant for the component of the nuclear spin magnetization vector perpendicular to the external magnetic field generated by the magnetic array 24.
[131] D is the molecular self-diffusion coefficient (also known as the "diffusion constant") of water molecules that was defined by A. Einstein in 1905 (A. Einstein in "Ann Physik", 17, page 549 (1905) ). Unlike Fick's law, no “gradient” is needed for its definition. Instead, one can think of a certain small volume of water molecules in a large volume. After waiting for a certain time t, several water molecules will "diffuse" out of this volume due to Brownian motion. The diffusion coefficient describes how quickly this process takes place. Einstein's equation describes the distance X for the water molecules traveling through Brownian motion: X2 = 2 •D •t.
[132] In NMR, this process can be measured using the NMR signal of water and the application of a magnetic field gradient. The Dof diffusion coefficient of water is altered by several anatomical details. For example, if there is a diffusion barrier, such as a cell membrane, D will be decreased. This can happen when an embryo in an egg is developed with surrounding biological structures such as blood vessels or something similar.
[133] Therefore, each of the 64 x 64 voxels has three parameter values T1, T2 and D associated with it, and paired combinations of parameters associated with the same voxel are illustrated in the off-diagonal diagrams in Figure 6. For example, in In the lower left diagram, the T1/D pairs for each voxel are indicated in a scatter plot, where the T1/D pairs corresponding to an egg that includes a male embryo are represented by a cross and the T1/D pairs corresponding to a egg that includes a female embryo are represented by a circle. Note that spatial information, ie where the location of an egg T1/D pair belongs, cannot be seen in this diagram, but of course it is available. Likewise, the diagram in the left column, middle row, shows the T1/T2 pairs, and the diagram in the middle column, third row, shows the T2/D pairs, equally represented by crosses and circles for male embryos. and females, respectively.
[134] The other three off-diagonal diagrams show the same combinations of parameters, but with the role of the horizontal and vertical axes interchanged and presented in a way, where sex is represented by the colors black and gray, which allows for better distinction. the areas associated with male/female embryos to the naked eye.
[135] The diagonal diagrams show histograms, in which, for each of the respective parameter binaries, the number of voxels found inside the compartment is counted. As can be seen in the three diagonal diagrams, for each of the three parameters T1, T2 and D, the histograms obtained for eggs containing male and female embryos differ. Although the diagonal diagrams in Figure 6 represent the histograms for measuring only a single egg of each sex, when calculating the average of the histograms for a plurality of eggs, as shown in Figures 7, 8 and 9, it can be seen that the deviations between histograms are really systematic.
[136] Figure 7 shows a mean histogram for the T1 values found in T1 images of 14 female and 12 male eggs. For both sexes, three peaks can be observed at T1, where the two peaks at the lowest T1 times are practically identical for both sexes. However, male embryos that contain eggs show a high peak of T1 at about 2250 ms, while female embryos that contain eggs show a high peak of T1 at longer T1 times, of about 2750 µs. The voxels in the T1 parameter image that display these long T1 values are precisely the voxels located at the top of the egg, where the embryo is located. This can also be seen at least qualitatively in Figure 5, where the eggs are shown in a horizontal configuration, and where the long T1 times are actually found in the upper portion of each egg. The sex-related difference in the high peak T1 observed in regions close to the embryo can therefore be used to determine the sex of the embryo.
[137] Figure 8 is similar to Figure 7, except that it shows a mean histogram for the T2 values. The histogram shows two T2 peaks, where the high T2 peak corresponds to voxels in the upper part of the egg and therefore close to the embryo. For eggs with a male embryo, the high T2 peak is again smaller and located at about 150 ms, while the high T2 peak for eggs with a female embryo is around 200 ms. The difference between T2 histograms, however, is not as pronounced as in the case of T1.
[138] Finally, Figure 9 shows a histogram similar to Figures 7 and 8 for the diffusion coefficient, which is seen to display three peaks. Again, the high diffusivity peak corresponds to voxels close to the location of the embryo, and its location is sex-dependent. In particular, the high diffusivity peak for eggs containing male embryos is around 1.75 mm2/s, whereas it is found around 2 mm2/s for eggs containing female embryos.
[139] As each of the parameters T1, T2 and D is sex sensitive, it can be used by the egg classification module 38 to determine the sex of the embryo contained in the egg. As indicated above, for each of these parameters, a representative parameter value for a region of interest within the egg, ie, near or at the location of the embryo, can be determined and then used in sex determination by the module. of classification 38.
[140] However, in preferred modalities, the classification module 38 receives images of entire parameters, like the images shown in Figure 5, and bases its determination on them. To this end, the classification module 38 is preferably a machine learning module. A machine learning module has the ability to learn from the data and then make predictions or decisions based on the data by building a model from sample inputs. For example, as suggested in Figure 6, in the five-dimensional parameter space covered by the parameters T1, T2, D, x coordinates and y coordinates, there are obviously distinctions between male and female embryonic eggs, and the pattern of this difference in that space of parameter can be very well recognized by machine learning algorithms.
[141] In machine learning and statistics, classification is the problem of identifying which set of categories, in this case male and female, a new observation belongs to, based on an observation training dataset, whose adherence of category is known. A "classifier" is an algorithm or machine that implements classification.
[142] Therefore, in preferred embodiments, classification module 38 is a machine learning module. Preferably, the values of the NMR parameters, or parameters derived therefrom, form feature values presented to the machine learning module as a feature vector.
[143] In preferred embodiments, classification module 38 is configured to determine embryo sex prediction using a linear classifier. Linear classifiers classify objects by making a classification decision based on the value of a linear combination of feature values. Suitable linear classifiers can be based on one or more linear regressions of least squares, nearest neighbors, logistic regression, and separation hyperplanes. The theory of linear classifiers is known to those skilled in the art of machine learning. For a detailed explanation of the aforementioned linear classifiers, reference is made to Hastie, T., Tibshirani, R. & Friedman, J. The Elements of Statistical Learning. Elements 1, (Springer, 2009). Another suitable linear classifier is the so-called perceptron algorithm, which is an algorithm for supervised learning of binary classifiers. One of the advantages of the perceptron algorithm is that it allows for online learning, as it processes elements of the training set one at a time. For more details on the perceptron algorithm, reference is made to Rosenblatt, F. The perceptron: A probabilistic model for information storage and organization in the brain. Psychol. Rev. 65, pages 386 to 408 (1958).
[144] In alternative embodiments, classification module 38 is configured to determine embryo sex prediction using a non-linear classifier. For embryo sex determination using the above NMR parameters as characteristic values, nonlinear classifiers based on polynomials per part, splines, kernel smoothing, tree-based methods, support vector machines, reinforcement methods, additive methods and ensemble, or graphic models, may be advantageously employed. Again, a detailed explanation of these non-linear classifiers can be taken from the above work by Hastie, which is incorporated herein by way of reference. A particularly suitable nonlinear classifier is based on the random forest method, which operates by building several decision trees at the time of training and produces the class which is called the decision tree class mode. For more details on the random forest method, reference is made to Criminisi, A. Decision Forests: The Unified Framework for Classification, Regression, Density Estimation, Manifold Learning and Semi-Supervised Learning. Foundations and Trends® in Computer Graphics and Vision 7, pages 81 to 227 (2011).
[145] In additionally alternative embodiments, classification module 38 is configured to determine embryo sex prediction using a deep learning algorithm. Deep learning is part of the broader family of machine learning methods, as discussed in this disclosure, and is based on representations of learning data as opposed to task-specific algorithms. For a review of suitable deep learning algorithms, reference is made to Y., L., Y., B. & G., H. Deep learning. Nature 521, pages 436 to 444 (2015). Deep learning algorithms particularly suitable for the purposes of the present invention are based on convolutional neural networks, as described in Le Cun, Y. et al. Handwritten Digit Recognition with a Back-Propagation Network. Adv. Neural Inf. Process. System Pages 396 to 404 (1990). doi: 10.1111/dsu.12130, based on recurrent neural networks, as described in Donahue, J. et al. Long-term Recurrent Convolutional Networks for Visual Recognition and Description. Cvpr 07-12 - June, pages 2625 to 2634 (2015), or based on short-term long memory networks, as described in Hochreiter, S. & Schmidhuber, J. Long Short-Term Memory. Neural Comp. 9, pages 1735 to 1780 (1997).
[146] The currently preferred implementation of a machine learning classifier for egg sexing is based on convolutional neural networks (CNN). Figure 14 schematically shows an overall architecture deployment. C parameter images of MxN pixels are shown on CNN. Convolutional filters C1 mxn, grouping and activation through the input are applied, resulting in a first layer of size M1xN1 with C1 channels. This process is repeated p times, until the features of interest from the parameter maps are extracted. Then, q neural layers of L channels are interconnected to give a final probability for a male or female.
[147] Deployments of these CNNs may include variations of convolution, clustering, non-linear activation, or architecture filters. Examples are, among others, AlexNet (Krizhevsky, A., Sutskever, I. & Hinton, ImageNet Classification with Deep Convolutional Neural Networks. Adv. Neural Inf. Process. Syst. 1-9 (2012). Doi: http:// /dx.doi.org/10.1016/j.protcy.2014.09.007), Overfeat (Sermanet, P. et al. OverFeat: Integrated Recognition, Localization and Detection using Convolutional Networks. arXiv Prepr. arXiv 1312.6229 (2013). doi: 10.1109/CVPR.2015.7299176), VGG (Simonyan, K. & Zisserman, A. Very Deep Convolutional Networks for Large-Scale Image Recognition. Int. Conf. Learn. Represent. 1-14 (2015). Doi: 10.1016/j. infsof.2008.09.005), Network-in-network (NiN) (Lin, M., Chen, Q. and Yan, S. Network In Network. arXiv Prepr.10 (2013). doi: 10.1109/ASRU.2015.7404828) , GoogLeNet and Inception (Szegedy, C. et al. Going deeper with convolutions. in Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition 7-12-June, 1-9 (2015)), ResNet (He, K ., Zhang, X., Ren, S. and Sun, J. Deep Resid Learning for Image Recognition. in 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 770 to 778 (2016). doi: 10.1109/CVPR.2016.90), SqueezeNet (Iandola, FN et al. SQUEEZENET:ALEXNET-LEVEL ACCURACY WITH 50XFEWERPARAMETERSAND <0.5MB MODEL SIZE. arXiv 1-5 (2016). doi: 10.1007/978-319-24553 -9 ) and ENet (Paszke, A., Chaurasia, A., Kim, S. and Culurciello, E. ENet: The Deep Neural Network Architecture for Real-Time Semantic Segmentation. arXiv 1-10 (2016)).
[148] According to the present understanding, the three NMR parameters T1, T2 and D above are the most relevant for sex prediction, and the most reliable results can be obtained if the classification module 38 bases its prediction on all the three parameters, in particular parameter images. However, the invention is not limited to this. In some modalities, a set of two or more NMR parameters is used, of which at least one is selected from said group consisting of a relaxation time T1, a relaxation time T2 and a diffusion coefficient. In such embodiments, the set of NMR parameters may further comprise one or more of the following parameters: a relaxation time T2*, a relaxation time Tip, and a spin density associated with one or more of the 1H, 13C, 23Na and nuclei. 31P.
[149] Additionally, or alternatively, the set of NMR parameters may additionally comprise one or more than one chemical shift signal from metabolites, in particular, water, lipids, amino acids, nucleic acids or hormones; a chemical shift selective transfer signal; and zero quantum coherence or multiple quantum coherence NMR signals.
[150] Rather than distinguishing between the sex of embryos in an egg 14, the apparatus 10 of Figure 1 can be operated to determine the fertility of an egg 14, as will be explained below with reference to Figures 10 to 12.
[151] Figure 10 shows a mean histogram of the diffusion coefficient observed across a plurality of fertile eggs (solid line) and infertile eggs (broken line). More precisely, diffusion coefficients were determined for each of the voxels of a diffusion coefficient image of the entire volume of a plurality of eggs. As seen in Figure 10, the histogram has very similar values around 2 mm2/s (which correspond to albumin), but diffusion coefficients around 1 mm2/s (found in regions within the yolk) are more often found. in infertile eggs than in infertile eggs. Consequently, this difference can be taken as a criterion to determine fertility.
[152] Figure 11 shows a scatter plot of pairs of histogram diffusion coefficient values at 1 mm2/s and 2 mm2/s for nine fertile and ten infertile eggs. As can be seen in Figure 11, in the simple scatter plot, all but one egg of each species are located on the corresponding side of a dashed separation line, which denotes a ratio of the histogram values at 2 and 1 mm2/ s, that the ratio is generally exceeded for fertile eggs and not reached for infertile eggs.
[153] Figure 12 shows a yolk NMR spectrum for fertile and infertile eggs. The spectrum shows a peak at about 1 ppm corresponding to fat and a peak at about 4.7 ppm corresponding to water. The inventors found that the ratio of peak height of fat to peak height of water is higher for infertile eggs than for fertile eggs. Consequently, based on this reason, a distinction between fertile and infertile eggs can also be made.
[154] In preferred modalities, the two fertility indicators, ie, the diffusion coefficient histogram format and the ratio of the fat to water peaks, can be combined to increase the reliability of the prediction. Note that comparing the histogram values at 1 mm2/s and 2 mm2/s is just one way to explore the characteristic shape of the diffusion coefficient histogram. In preferred embodiments, the entire diffusion coefficient histogram can be presented to a machine learning algorithm, which automatically learns to distinguish between diffusion coefficient histograms corresponding to fertile and infertile eggs. Figure 11 indicates that there is enough fertility-related information in the diffusivity histogram to make the correct distinction, which can be properly accounted for by a machine learning module, such as a properly configured egg grading module 38.
[155] Similarly, although the ratio of fat to water peaks in the spectrum in Figure 12 is only one way of distinguishing egg fertility based on spectrum, in alternative modalities, the entire spectrum can be presented to a module of machine learning, such as a properly configured egg sorting module 38 that, after sufficient training, could distinguish between fertile and infertile eggs on the basis of spectrum.
[156] Note further that the distinction between fertile and infertile eggs 14 can be made with the same apparatus 10, as shown in Figure 1, the only difference being the NMR measurement protocol, which is provided by the central controller 44 at components of the 18 NMR apparatus, how to measure the parameters of interest, and the algorithm employed by the egg classification module 38.
[157] Figure 13 shows a simplified device 10' that has the same functionality as the device 10 of Figure 1, except that it was not designed for automated determination of sex or high-yielding fertility. Apparatus 10' comprises an NMR apparatus 18, which basically includes the same type of components as apparatus 10 of Figure 1, but is designed to measure only a single egg. The simplified apparatus of Figure 13 does not include a transport device 12 and also does not have an egg separating device 40. Instead, the output of the egg sorting module 38 is presented to a user interface 60. This simplified version it can be used for laboratory scale analysis rather than industrial analysis. While preferred embodiments of the invention are aimed at high-throughput automated methods and apparatus, which include a transport device 12 and an egg separating device 40, and are aimed at carrying out measurements on a plurality of eggs 14 in parallel, the invention is not limits this, and all of the modalities described in this document can also apply to apparatus that do not require the transport device 12 and the egg separating device 40.
[158] Although an exemplary preferred embodiment is shown and specified in detail in the drawings and in the above descriptive report, it should be seen as purely exemplary and not limiting of the invention. It is noted, in this regard, that only the exemplary preferred embodiment is shown and specified, and all variations and modifications must be protected that, currently or in the future, are within the scope of protection of the invention, as defined in the claims. LIST REFERENCES10 Apparatus for Non-Invasive Determination of the Sex of an Embryo or the Fertility of an Egg12 Transport Device14 Egg16 Tray18 NMR Apparatus20 Conveyor Belt22 Transport Controller24 Magnet Arrangement26 Gradient Coils28 Gradient Controller30 RF Coil Arrangement RF32 Coil RF34 Controller RF36 Detector Image Reconstruction Module38 Egg Sorting Module40 Egg Separating Device42 Egg Separating Device Suction Container 4044 Central Controller45 Data Channel46 Shipping Direction48 Tray Cavity 1650 Antenna Section52 Tuning Capacitor54 Preamplifier56 Transmission Lines58 are RMN60 Pectrometer User Interface

权利要求:
Claims (15)
[0001]
1. Method of automated non-invasive determination of the sex of an embryo from a bird egg (14), the method characterized in that it comprises the following steps: transporting a plurality of bird eggs (14), sequentially or in parallel, to an NMR apparatus (18), subject the bird eggs (14) to an NMR measurement and determine, for each of said eggs (14), one or more NMR parameter value images of a region of the egg. , wherein the NMR parameter values are associated with the corresponding pixels or voxels of the image, wherein the one or more NMR parameters are selected from the group consisting of a relaxation time T1, a relaxation time T2 and a coefficient broadcast, forwarding the one or more NMR parameter value images, or parameter images derived therefrom, to a classification module (38), said classification module (38) being configured to determine on the basis of in the one or more param value images NMR meters or images of parameters derived therefrom, predicting the sex of the associated egg embryo (14), and transporting said plurality of bird eggs (14) out of said NMR apparatus (18) and separating the eggs ( 14) according to the sex prediction provided by said classification module (38).
[0002]
2. Method according to claim 1, characterized in that said one or more NMR parameters comprise a set of two or more NMR parameters, of which at least one is selected from said group consisting of a relaxation time T1, a relaxation time T2 and a diffusion coefficient, wherein said set of NMR parameters preferably further comprises one or more of the following parameters: a relaxation time T2*, a relaxation time Tip and a spin density associated with one or more of the 1H, 13C, 23Na and 31P nuclei, and/or wherein said set of NMR parameters preferably further comprises one or more of a chemical shift signal of metabolites, in particular water, lipids , amino acids, nucleic acids or hormones; a chemical shift selective transfer signal; and zero quantum coherence or multiple quantum coherence NMR signals.
[0003]
3. Method according to claim 1 or 2, characterized in that said classification module (38) is a machine learning module.
[0004]
4. Method according to any one of claims 1 to 3, characterized in that said classification module is configured to determine the prediction of the sex of the embryo using a linear classifier, in particular, a linear classifier based on one or more linear regressions of least squares, nearest neighbors, logistic regression, separation hyperplanes or perceptrons, or wherein said classification module (38) is configured to determine the prediction of embryo sex using a nonlinear classifier, in particular, a nonlinear classifier based on piecewise polynomials, splines, kernel smoothing, tree-based methods, support vector machines, random forest, reinforcement, additive and ensemble methods or graph models, or where the said classification module (38) is configured to determine the prediction of the sex of the embryo using a deep learning algorithm, in particular, a learning algorithm. deep learning based on convolutional neural networks, recurrent neural networks or short-term long memory networks.
[0005]
5. Method according to any one of claims 1 to 4, characterized in that the measurement by NMR comprises imaging by NMR, in which an NMR imaging plan is arranged so as to intersect with the embryo site, and/or that said method is carried out before the eighth day of reproduction, preferably on the fifth day of reproduction.
[0006]
6. Method of automated non-invasive determination of the fertility of a bird egg (14), the method characterized in that it comprises the following steps: transporting a plurality of bird eggs (14), sequentially or in parallel, to an apparatus of NMR (18), subject the bird eggs (14) to an NMR measurement, to thus determine, for each of said eggs (14), one or both of a histogram of diffusion coefficients at various locations of the egg, - an NMR spectrum of the yolk which includes peaks corresponding to water and fat, determine a fertility prediction based on the histogram format of the diffusion coefficients and/or the NMR spectrum, and transport said plurality of bird eggs (14) out of said NMR apparatus (18) and sort the eggs (14) according to the fertility prediction.
[0007]
7. Method according to claim 6, characterized in that determining fertility based on the histogram format of the diffusion coefficients comprises comparing the frequency of occurrence of at least two different diffusion coefficients or two ranges of diffusion coefficients ,wherein said at least two different diffusion coefficients, or the centers of said at least two diffusion coefficient ranges, are preferably separated between 0.5 and 2.5 mm2/s, preferably between 0.75 and 1.5 mm2/s, and/or that of said at least two different diffusion coefficients, or of the centers of said at least two diffusion coefficient ranges, one is preferably located in the range of 0.6 to 1.3 mm2 /s preferably in a range of 0.7 to 1.2 mm2/s, and the other is located in a range of 1.5 to 2.5 mm2/s, preferably in the range of 1.7 at 2.3 mm2/s, and/or that said various locations in the egg are preferably evenly distributed in the egg and, in particular, correspond to voxels of a diffusion coefficient image, and/or wherein said step of determining a fertility prediction based on the NMR spectrum preferably comprises determining said fertility prediction based on the ratio of peaks corresponding to water and fat in said NMR spectrum.
[0008]
8. Method according to any one of claims 1 to 7, characterized in that said eggs (14) are arranged in a regular pattern, in particular in a matrix configuration, on a tray (16) during the said transport and measurement by NMR, wherein preferably the number of eggs (14) arranged in said tray (16) is at least 36, preferably at least 50 and most preferably at least 120, and/or which said NMR apparatus (18) preferably comprises an array (30) of RF coils (30a) to apply RF magnetic fields to eggs (14) located in the tray (16) and/or to detect NMR signals, wherein said arrangement (30) of RF coils (30a) comprises one or more of a plurality of coils (30a) arranged in a plane located above the tray (16) loaded with eggs (14), when transported to the delivery apparatus. NMR (18), - a plurality of coils (30a) arranged in a plane located below the tray (16) loaded with eggs (14) when transported to the NMR apparatus (18), - a plurality of coils (30a) arranged in vertical planes, extending between rows of eggs (14) in the tray (16), when transported to the NMR apparatus (18), whose rows extend parallel to the direction of transport of the tray (16), into and out of the NMR apparatus (18), in which preferably, in the case of the plurality of coils (30a) arranged in a localized plane above or below the tray (16) loaded with eggs (14), the ratio of the number of coils (30a) to the number of eggs (14) arranged in said tray (16) is between 1:1 and 1:25, of preferably between 1:1 and 1:16, and most preferably between 1:1 and 1:5, and/or wherein said NMR apparatus (18) preferably comprises an array (30) of RF coils ( 30a) to apply RF magnetic fields to the eggs (14) located in the tray (16) and/or to detect NMR signals, said arrangement (30) of RF coils (30a) being integrated or connected. attached to said tray (16), wherein the tray (16) preferably comprises a plurality of cavities (48) or pockets for receiving a corresponding egg (14) and wherein a number of spools (30a) is associated with each of the said cavities (48) or pockets, wherein said number of spools (30a) per cavity (48) or pocket is at least 1, preferably at least 2, most preferably at least 3 and most preferably , at least 4 and/or wherein at least some of said spools (30a) are disposed vertically with respect to the main plane of the tray (16), or at an angle of at least 50°, preferably of at least 75 ° and, most preferably, at least 80° with respect to the main plane of the tray (16).
[0009]
9. Apparatus (10) for automated non-invasive sex determination of an embryo of a bird egg (14), the apparatus characterized in that it comprises: an NMR apparatus (18), a transport device (12) to transport a plurality of bird eggs (14), sequentially or in parallel, in said NMR apparatus (18) and outside of said NMR apparatus (18), wherein said NMR apparatus (18) is configured to submit the bird eggs (14) to an NMR measurement, to thereby determine, for each of said eggs (14), one or more NMR parameter value images of a region of the egg (14), wherein NMR parameter values are associated with the corresponding pixels or voxels in the image, wherein the one or more NMR parameters are selected from the group consisting of a relaxation time T1, a relaxation time T2 and a diffusion coefficient, wherein said apparatus (10) further comprises a classification module (38) configured to receive a one or more NMR parameter value images, or parameter images derived therefrom, said classification module (38) being configured to determine, based on the one or more NMR parameter value images or parameter images derived therefrom, a sex prediction of the associated egg embryo (14), and an egg separating device (40) for separating the eggs (14) in accordance with the sex prediction provided by said classification module (38).
[0010]
10. Apparatus (10) according to claim 9, characterized in that said one or more NMR parameters comprise a set of two or more NMR parameters, of which at least one is selected from said group that it consists of a relaxation time T1, a relaxation time T2 and a diffusion coefficient, wherein said set of NMR parameters preferably further comprises one or more of the following parameters: a relaxation time T2*, a relaxation time Tip and a spin density associated with one or more 1H, 13C, 23Na and 31P nuclei, and/or wherein said set of NMR parameters preferably further comprises one or more of a chemical shift signal from metabolites, in particular, water, lipids, amino acids, nucleic acids or hormones; a chemical shift selective transfer signal; and zero quantum coherence or multiple quantum coherence NMR signals.
[0011]
11. Apparatus (10) according to claim 9 or 10, characterized in that said classification module (38) is a machine learning module, wherein said classification module (38) is preferably configured to determine the prediction of the sex of the embryo using a linear classifier, in particular, a linear classifier based on one or more of linear regression by least square, nearest neighbors, logistic regression, separation hyperplanes or perceptrons, or in which said module classification (38) is preferably configured to determine the prediction of embryo sex using a non-linear classifier, in particular a non-linear classifier based on piecewise polynomials, splines, kernel smoothing, tree-based methods, support vectors, random forest, reinforcement, additive and ensemble methods or graphics models, or wherein said classification module (38) is configured preferentially primarily to determine embryo sex prediction using a deep learning algorithm, in particular, a deep learning algorithm based on convolutional neural networks, recurrent neural networks or short-term long memory networks, or in which the classification module (38) is preferably configured to determine the prediction of the sex of the embryo based on a comparison with parameter values stored in a database.
[0012]
12. Apparatus (10) for automated non-invasive determination of the fertility of a bird egg (14), the apparatus characterized in that it comprises: an NMR apparatus (18), a transport device (12) for transporting a plurality of bird eggs (14), sequentially or in parallel, in said NMR apparatus (18) and outside of said NMR apparatus (18), wherein said NMR apparatus (18) is configured to subject eggs to bird (14) to an NMR measurement, to thereby determine, for each of said eggs (14), one or both of - a histogram of scattering coefficients at various locations in the egg, - an NMR spectrum of the yolk which includes peaks corresponding to water and fat, wherein said apparatus (10) is further configured to determine a fertility prediction based on the histogram format of the diffusion coefficients and/or the NMR spectrum, wherein said apparatus ( 10) further comprises an egg separating device (40) for separating the eggs (14 ) according to the fertility prediction.
[0013]
13. Apparatus (10) according to claim 12, characterized in that determining fertility based on the histogram format of the diffusion coefficients, by the apparatus (10), comprises comparing the frequency of occurrence of at least two coefficients of different diffusion or two bands of diffusion coefficients, wherein said at least two different diffusion coefficients, or the centers of said at least two bands of diffusion coefficients, are preferably separated between 0.5 and 2.5 mm2/s preferably between 0.75 and 1.5 mm2/s, and/or that of said at least two different diffusion coefficients, or centers of said at least two diffusion coefficient ranges, one is preferably located in a range from 0.6 to 1.3 mm2/s, preferably in the range from 0.7 to 1.2 mm2/s, and the other is located in the range from 1.5 to 2.5 mm2/s, preferably , in the range of 1.7 to 2.3 mm2/s, and/or wherein said various locations in the egg are preferably di. distributed uniformly in the egg and, in particular, correspond to voxels of a diffusion coefficient image, and/or wherein said apparatus (10) is preferably configured to determine a fertility prediction based on the NMR spectrum based on the ratio of peaks corresponding to water and fat in said NMR spectrum.
[0014]
14. Apparatus (10) according to claim 12 or 13, characterized in that it further comprises a tray (16), on which said eggs (14) can be arranged in a regular pattern, in particular in a matrix configuration, during said transport and measurement by NMR, wherein the number of eggs (14) that can be arranged on said tray (16) is preferably at least 36, more preferably at least 50 and most preferably at least 120.
[0015]
15. Apparatus (10) according to claim 14, characterized in that said NMR apparatus (18) comprises an arrangement (30) of RF coils (30a) for applying RF magnetic fields to eggs (14) located on the tray (16) and/or for detecting NMR signals, said array (30) of RF coils (30a) comprising one or more of a plurality of coils (30a) arranged in a plane located above the tray (16) loaded with eggs (14), when transported to the NMR apparatus (18), - a plurality of coils (30a) arranged in a plane located below the tray (16) loaded with eggs (14), when transported for the NMR apparatus (18), - a plurality of coils (30a) arranged in vertical planes extending between rows of eggs (14) in the tray (16), when transported to the NMR apparatus (18), whose rows extend parallel to the transport direction of the tray (16), into and out of the NMR device (18), in which case of the plurality of coils (30a) arranged in a plane located above or below the tray (16) loaded with eggs (14), the ratio of the number of coils (30a) to the number of eggs (14) disposed on said tray (16 ) is preferably between 1:1 and 1:25, more preferably between 1:1 and 1:16, and most preferably between 1:1 and 1:5, and/or wherein said NMR device (18 ) comprises an array (30) of RF coils (30a) to apply RF magnetic fields to eggs (14) located in the tray (16) and/or to detect NMR signals, said array (30) of coils of RF (30a) is integrated or connected to said tray (16), wherein the tray (16) preferably comprises a plurality of cavities (48) or pockets for receiving a corresponding egg (14) and wherein a number of spools ( 30a) is associated with each of said cavities (48) or pockets, wherein said number of spools (30a) per cavity (48) or pocket is at least 2, preferably at least 3 and with maximum p reference, at least 4 and/or wherein at least some of said spools (30a) are disposed vertically with respect to the main plane of the tray (16), or at an angle of at least 50°, preferably of at least 75° and most preferably at least 80°, relative to the main plane of the tray (16).

类似技术:

---

公开号 | 公开日 | 专利标题

---

BR112020008747B1|2021-07-27|AUTOMATED NON-INVASIVE DETERMINATION OF THE SEX OF AN EMBRYO AND THE FERTILITY OF A BIRD EGG.

---

US6029080A|2000-02-22|Method and apparatus for avian pre-hatch sex determination

---

ES2289011T3|2008-02-01|RM PROCEDURE TO ANALYZE SEEDS.

---

AU2001266865C1|2004-08-19|Methods and apparatus for non-invasively identifying conditions of eggs via multi-wavelength spectral comparison

---

US20130044210A1|2013-02-21|Hyperspectral identification of egg fertility and gender

---

JP6723597B2|2020-07-15|Non-destructive inspection device for egg and egg inspection program used therefor

---

WO2010103111A1|2010-09-16|Method for avian sex determination

---

BR112020001373A2|2020-10-06|system and method for noninvasive determination of egg properties

---

Smith et al.2008|Fertility and embryo development of broiler hatching eggs evaluated with a hyperspectral imaging and predictive modeling system

---

CN109862783B|2021-11-02|Hatching egg inspection system and hatching egg inspection program

---

Celik et al.2017|Estimation of egg weight from some external and internal quality characteristics in quail by using various data mining algorithms

---

ES2769753T3|2020-06-29|Identification system for honeycomb egg trays and associated procedure

---

Mbajiorgu2011|Effect of hatching egg size on hatchability and chick hatch-weight of indigenous Venda chickens

---

JP2020204570A|2020-12-24|Hatching egg inspection device, method for specification, and specification program

---

BR112017027842B1|2021-10-05|EGG TRAY IDENTIFICATION SYSTEM AND METHOD

---

Giancardo et al.2012|Automatic tractography analysis through sparse networks in case-control studies

---

Milisits et al.2010|Possibilities and Deliminations In Vivo Determination of Egg Composition

---

Donkó et al.2010|Predictability of Egg Content in Hen's Eggs by Means of Cross Sectional Image Analysis

---

Mbajiorgu2011|EFFECT OF HA EFFECT OF HATCHING EGG SIZE ON HA CHING EGG SIZE ON HA CHING EGG SIZE ON HATCHABILITY CHABILITY AND CHICK HA AND CHICK HATCH-WEIGHT OF INDIGENOUS VENDA CHICKENS WEIGHT OF INDIGENOUS VENDA CHICKENS

同族专利:

---

公开号 | 公开日

---

SI3523649T1|2021-03-31|

---

DK3523649T3|2021-02-15|

---

HUE052953T2|2021-05-28|

---

US11116184B2|2021-09-14|

---

BR112020009533A2|2020-11-03|

---

BR112020008747A2|2021-02-23|

---

EP3798659A1|2021-03-31|

---

WO2019092265A1|2019-05-16|

---

EP3545326A1|2019-10-02|

---

CN111344587A|2020-06-26|

---

US20200281166A1|2020-09-10|

---

WO2019092267A1|2019-05-16|

---

PL3545326T3|2021-04-06|

---

HUE052046T2|2021-04-28|

---

LT3523649T|2021-02-25|

---

LT3545326T|2020-12-10|

---

EP3523649A1|2019-08-14|

---

DK3545326T3|2020-12-07|

---

SI3545326T1|2021-01-29|

---

RU2739153C1|2020-12-21|

---

EP3483620A1|2019-05-15|

---

EP3483619A1|2019-05-15|

---

ES2829964T3|2021-06-02|

---

CN111386462A|2020-07-07|

---

US20200348248A1|2020-11-05|

---

EP3545326B1|2020-09-09|

---

ES2854001T3|2021-09-20|

---

EP3523649B1|2020-11-11|

---

RU2739896C1|2020-12-29|

---

BR112020009533B1|2021-03-23|

---

PT3523649T|2021-01-27|

---

PT3545326T|2020-11-26|

---

US11122778B2|2021-09-21|

引用文献:

---

公开号 | 申请日 | 公开日 | 申请人 | 专利标题

---

---

EP1021717B1|1996-10-04|2002-04-10|Embrex Inc.|Method of sorting birds in ovo|

---

US5745228A|1997-01-17|1998-04-28|Embrex, Inc.|Method and apparatus for distinguishing live from infertile poultry eggs|

---

NL1006504C2|1997-07-07|1999-01-08|Instituut Voor Dierhouderij En|Method and device for selecting eggs.|

---

US6029080A|1997-07-09|2000-02-22|Reynnells; Richard D.|Method and apparatus for avian pre-hatch sex determination|

---

US6234320B1|1999-05-11|2001-05-22|Embrex, Inc.|Method and apparatus for selectively classifying poultry eggs|

---

US6535277B2|2000-12-20|2003-03-18|Embrex, Inc.|Methods and apparatus for non-invasively identifying conditions of eggs via multi-wavelength spectral comparison|

---

US7041439B2|2001-04-17|2006-05-09|Embrex, Inc.|Methods and apparatus for selectively processing eggs having identified characteristics|

---

JP4128791B2|2002-03-27|2008-07-30|株式会社堀内|Sexual differentiation technology for sperm eggs|

---

WO2006057081A1|2004-11-24|2006-06-01|Mitsuyo Okazaki|Method for analyzing lipoproteins|

---

US7880467B2|2005-06-09|2011-02-01|Aspect Magnet Technologies Ltd.|Packed array of MRI/NMR devices and an MRI/NMR method of analyzing adjacent lines of goods simultaneously|

---

CN101273276A|2005-09-22|2008-09-24|威斯康星校友研究基金会|Backprojection reconstruction method for ct imaging|

---

US7950439B2|2005-11-28|2011-05-31|Victor Ray Anderson|Combination-action slide and hinge swinging door|

---

US20070162992A1|2006-01-09|2007-07-12|Mcgill University|Metabolomic determination in assisted reproductive technology|

---

BRPI0708212A2|2006-02-23|2011-05-17|Ovagen Internat Ltd|improved methods of producing specified germ-free bird and bird eggs|

---

WO2009134820A2|2008-04-28|2009-11-05|Cornell University|Tool for accurate quantification in molecular mri|

---

KR20100101526A|2009-03-09|2010-09-17|삼성전자주식회사|Strobo thin film chemical analysis apparatus and assay method using the same|

---

WO2010103111A1|2009-03-13|2010-09-16|Van De Ven Beheer B.V.|Method for avian sex determination|

---

WO2010150265A2|2009-06-25|2010-12-29|Yissum Research Development Company Of The Hebrew University Of Jerusalem Ltd.|Hyperspectral identification of egg fertility and gender|

---

CN101672839B|2009-09-25|2012-11-14|南京农业大学|Device and method for detecting hatching egg incubation quality based on computer vision|

---

JP2011142866A|2010-01-15|2011-07-28|Ryosuke Taniguchi|Method for identifying sex of fertilized egg based on outline thereof|

---

DE102010006161B3|2010-01-21|2011-01-13|Technische Universität Dresden|Determining gender of fertilized and non-incubated bird eggs, where egg has yolk surrounded by egg shell and embryonic disk arranged to yolk, by guiding probe into egg shell through hole in direction to the disk with embryonic disk cells|

---

NZ602522A|2010-05-21|2014-06-27|Xl Protein Gmbh|Biosynthetic proline/alanine random coil polypeptides and their uses|

---

US9710730B2|2011-02-11|2017-07-18|Microsoft Technology Licensing, Llc|Image registration|

---

TR201901409T4|2012-07-30|2019-02-21|In Ovo B V|Determination of the sex, viability and / or developmental stage of bird embryos in ovo.|

---

DE102013205426A1|2013-03-27|2014-10-16|Siemens Aktiengesellschaft|Method for determining the sex of an embryo in an egg|

---

US9395346B2|2013-11-18|2016-07-19|Zoetis Services Llc|Non-contact egg identification system for determining egg viability, and associated method|

---

CN107072595B|2013-12-31|2021-11-26|威斯康星州医药大学股份有限公司|Adaptive re-planning based on multi-modality imaging|

---

JP6671348B2|2014-05-05|2020-03-25|アトムワイズ，インコーポレイテッド|Binding affinity prediction system and method|

---

DE102014010150A1|2014-07-04|2016-01-07|Technische Universität Dresden|Method and apparatus for Raman spectroscopic in ovo sexing of fertilized and incubated bird eggs|

---

FR3023617B1|2014-07-11|2018-06-15|Tronico|NON - INVASIVE DEVICE FOR DETERMINING THE FERTILITY AND / OR SEX OF AN EGG AND CORRESPONDING METHOD.|

---

US9459201B2|2014-09-29|2016-10-04|Zyomed Corp.|Systems and methods for noninvasive blood glucose and other analyte detection and measurement using collision computing|

---

CN104316473A|2014-10-28|2015-01-28|南京农业大学|Gender determination method for chicken hatching egg incubation early embryo based on hyperspectral image|

---

US10478117B2|2015-04-08|2019-11-19|Ramot At Tel-Aviv University Ltd.|Early detection of reduced bone formation with an NMR scanner|

---

EP3362783A1|2015-10-12|2018-08-22|MM Limited|A foodstuff item characteristic mri detection system|

---

DE102016004051B3|2016-04-04|2017-07-27|Technische Universität Dresden|Method and device for optical in-ovo sexing of fertilized and incubated bird eggs|

---

DE102016206713B4|2016-04-20|2017-11-02|Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V.|System for performing spectroscopic echo-planar imaging |

---

CN106645605B|2016-09-05|2019-01-08|中国农业大学|Birds fertile egg detection method based on machine vision|

---

CN106485251B|2016-10-08|2019-12-24|天津工业大学|Egg embryo classification based on deep learning|

---

CN107041325B|2017-04-06|2020-06-05|四川农业大学|Nondestructive gender determination method for avian embryo eggs|

---

EP3483619A1|2017-11-13|2019-05-15|Technische Universität München|Automated noninvasive determining the sex of an embryo of and the fertility of a bird's egg|EP3483619A1|2017-11-13|2019-05-15|Technische Universität München|Automated noninvasive determining the sex of an embryo of and the fertility of a bird's egg|

---

WO2020262557A1|2019-06-28|2020-12-30|四国計測工業株式会社|Egg inspection device, egg inspection program, and egg inspection method|

---

CN110516525B|2019-07-01|2021-10-08|杭州电子科技大学|SAR image target recognition method based on GAN and SVM|

---

CN110771533B|2019-09-29|2022-03-01|北京农业智能装备技术研究中心|System and method for detecting gender of embryonated egg|

---

CN111193920B|2019-12-31|2020-12-18|重庆特斯联智慧科技股份有限公司|Video picture three-dimensional splicing method and system based on deep learning network|

---

CN111345267B|2020-04-08|2021-11-19|山东省农业科学院家禽研究所（山东省无特定病原鸡研究中心）|Breeding method of hybrid line of special laying hens with feather color double-self-distinguishing male and female black-feather pink shells and high yield|

---

CN111517092B|2020-06-03|2021-06-11|太原理工大学|Transmission belt tearing detection method|

---

CN112517424A|2020-11-28|2021-03-19|唐少阳|Sorting and recycling device for defective eggs in animal breeding and operation method thereof|

法律状态:
2021-06-08| B09A| Decision: intention to grant [chapter 9.1 patent gazette]|

---

2021-07-27| B16A| Patent or certificate of addition of invention granted|Free format text: PRAZO DE VALIDADE: 20 (VINTE) ANOS CONTADOS A PARTIR DE 13/11/2018, OBSERVADAS AS CONDICOES LEGAIS. |

优先权:

---

申请号 | 申请日 | 专利标题

---

EP17201373.2A|EP3483619A1|2017-11-13|2017-11-13|Automated noninvasive determining the sex of an embryo of and the fertility of a bird's egg|

---

EP17201373.2|2017-11-13|

---

PCT/EP2018/081019|WO2019092265A1|2017-11-13|2018-11-13|Automated noninvasive determining the sex of an embryo of and the fertility of a bird's egg|

---